neuromuscular synaptogenesis: coordinating partners...

During embryonic development in vertebrates, motor neuron axons grow long distances from the spinal cord to reach their distal targets1. They form the link between the CNS and the rest of the body — particularly the striated muscles that effect voluntary movements. The direction of this long-distance axonal travel towards specific muscles to be innervated requires well-orchestrated interactions among axons, axonal Schwann cells and environmental cues1,2. Owing to its large size, accessibility and simple organization, the neuromuscular junction (NMJ), which ena-bles the nervous system to control the muscular system, has been used extensively to study synapse formation3,4. However, this apparent simplicity belies the complex pro-cesses that take place during synapse formation and that are necessary for NMJ function. Indeed, the formation of perfectly aligned presynaptic assemblies and postsynap-tic compartments requires an intricate interplay between molecular signals. Ultimately, the two elements become tuned functionally to one another, with the level of pre-synaptic transmitter release adapted to the capacity of postsynaptic activation.

Synapse formation is followed by another major step, synapse maturation, which leads to the stabilization of all of the synaptic elements to form a reliable synapse. It has long been acknowledged that the maturations of pre- and postsynaptic elements at the NMJ are inter-dependent; however, until recently, the elaborate inter-actions between these two compartments were poorly

understood. Presynaptic maturation was previously assumed to be controlled by molecules that are released by the postsynaptic apparatus, and vice versa5–8. However, recent advances in molecular analysis (for instance, improvements in fusion proteins and cell-specific gene deletion) have revealed a more complex scenario, wherein some molecules regulate both pre- and postsynaptic organization. Thus, it is possible that the formation and maturation of pre- and postsynaptic compartments are in fact tightly coordinated by shared molecular mecha-nisms. Moreover, although NMJ formation was originally thought to be regulated by only pre- and postsynaptic mechanisms, more recent evidence has implicated a role for glial molecular mechanisms in coordinating the maturation of the two synaptic elements. Indeed, it was reported that perisynaptic Schwann cells (PSCs; also known as terminal Schwann cells) produce and release synaptogenic factors9–14, and their own maturation is modulated by pre- and postsynaptic molecules15–20. This tri-directional interaction is also required for key activity-dependent changes that are involved in NMJ maturation and is particularly important during synaptic competi-tion and elimination. Indeed, PSCs detect activity from competing synapses18 and are responsible for debris clearance and the elimination of superfluous nerve ter-minals21,22. Thus, PSCs could represent a link between the activity-dependent and molecularly regulated changes that are essential for the maturation of the NMJ.

Neuromuscular junction(NMJ). A unitary functional structure composed of a single axon terminal innervating a muscle fibre. The presynaptic terminal is covered by specialized glial cells called perisynaptic Schwann cells.

Perisynaptic Schwann cells(PSCs). Non-myelinating glial cells at the neuromuscular junction. They originate from the neural crest but differ structurally and phenotypically from axonal myelinating or axonal non-myelinating Schwann cells.

Neuromuscular synaptogenesis: coordinating partners with multiple functionsHoussam Darabid, Anna P. Perez-Gonzalez and Richard Robitaille

Abstract | The formation of highly efficient and reliable synapses at the neuromuscular junction (NMJ) relies on dynamic molecular interactions. Studies of the development and maturation of the NMJ have focused on events that are dependent on synaptic activity and that require the coordinated actions of nerve- and muscle-derived molecules with different targets and effects. More recently, perisynaptic Schwann cells — the glial cells at NMJs — have become an important focus of research. These glia concomitantly contribute to pre- and postsynaptic maturation while undergoing maturation themselves. Thus, an intricate ‘danse à trois’ regulates the maturation of the NMJ to form a highly efficient communication unit, in which fine glial processes lie in close proximity to a highly concentrated population of postsynaptic receptors and perfectly aligned presynaptic release sites.

Groupe de Recherche sur le Système Nerveux Central, Département de Neurosciences, Université de Montréal, PO BOX 6128, Station Centre-Ville, Montréal, Québec H3C 3J7, Canada.Correspondence to R.R. e-mail: [email protected]:10.1038/nrn3821 Published online 1 October 2014

REVIEWS

NATURE REVIEWS | NEUROSCIENCE VOLUME 15 | NOVEMBER 2014 | 703

© 2014 Macmillan Publishers Limited. All rights reserved

mailto:richard.robitaille%40umontreal.ca?subject=Nature%20Reviews

mailto:richard.robitaille%40umontreal.ca?subject=Nature%20Reviews

Active zonesAreas on the surface of the presynaptic terminal that are characterized by their electron-dense appearance owing to the high concentration of proteins involved in Ca2+-dependent synaptic- vesicle exocytosis and recycling.

Motor columnsGroups of motor neurons that innervate selective sets of muscles.

Pre-patterningA nerve-independent phenomenon that occurs prior to the arrival of motor axons whereby acetylcholine receptors cluster in the central region of the muscle fibres (along the longitudinal axis), purportedly defining the location of nerve–muscle contact.

AChR clusteringThe gathering of acetylcholine receptors (AChRs), which is regulated by molecular mechanisms. It is one of the initial steps of synapse maturation.

In this Review, we discuss the growing evidence of a tri-directional molecular and activity-dependent crosstalk among presynaptic terminals, postsynaptic muscle fibres and PSCs at the developing NMJ. This coordination struc-tures and organizes the presynaptic release machinery and postsynaptic receptors, and thus enables the tight assembly of an efficient and reliable communication unit.

Molecular mechanisms of NMJ developmentNMJ formation, which is initiated during embryonic development (around embryonic day 12 (E12) to E13.5 in rodents)5,7,23,24, is preceded by the guidance of the axon towards the muscle fibre that is to be innervated and involves accurate molecular signalling to ensure the formation of precise nerve–muscle contacts. A period of maturation follows that leads to morphological and functional changes to the pre- and postsynaptic elements. Maturation of the NMJ depends on molecular mecha-nisms that elaborate the complex pre- and postsynaptic organization and that reduce the number of axonal termi-nals at a single NMJ until only one remains; these mecha-nisms are necessary for efficient synaptic communication. Presynaptic terminals must express synaptic proteins to form active zones, whereas postsynaptic elements express high densities of aggregated receptors at the crests of post-synaptic folds. Together, it is the ‘synaptic elements’ — namely, the presynaptic nerve terminal, the postsynaptic muscle fibre and the associated perisynaptic glia — that collectively coordinate the maturation of the NMJ.

Pathfinding, guidance and pre-patterning. During embryonic development, pools of motor neurons in the spinal cord form organized motor columns (already present by E6–7 in chick embryos)25–27, which can be classified according to their location (rostrocaudal, medio lateral or dorsoventral) within the spinal cord27,28. This columnar organization seems to be crucial for estab-lishing the proper inputs of each class of motor neuron to their specific effector targets to control complex loco-motor behaviour27,29,30. This well-structured organiza-tion suggests that elaborate mechanisms are required to guide the right group of motor axons to their specific final target. It is now recognized that pathfinding is not controlled by the target muscle per se31 but instead by guidance cues along the path of the motor neuron27,32, and that this process is dependent on neuronal activity33. These mechanisms, and the wide range of genes and tran-scription factors involved, have been discussed elsewhere (reviewed in REFS 1,27).

When motor axons reach their final target (the timing of which varies according to the muscle and its location; around E12.5 for the mouse diaphragm)23, muscle fibres already show a certain level of maturation, as indicated by the presence of clusters of ionotropic acetylcholine receptors (AChRs) on their surface23,34,35. The presence of AChR clusters on the fibres prior to the arrival of the motor axon indicates that these AChRs are not directly linked to synaptic activity. This has raised the question of whether the pre-patterning (that is, nerve-independent clustering) of AChRs dictates the location of NMJs on muscle fibres23,34,36–38. Although initial work in cell

culture has shown that NMJs do not form preferentially at pre-existing AChR clusters39,40, evidence obtained from mammals and other vertebrates suggests that AChR pre-patterns have a role in defining the location of nerve–muscle contacts37,38,41,42.

Upon contact with muscles, motor axons innervate muscle fibres extensively to generate polyinnervated NMJs and large motor units43–45. Although NMJs are functional at this stage, the motor neurons are not effi-cient enough to induce muscle contractions in a refined and precise way; this efficiency requires NMJ matura-tion. Even though pre- and postsynaptic maturation are discussed separately below, the two processes are inter-dependent and involve molecules that are derived from, and act on, all three synaptic elements.

Postsynaptic maturation. The postsynaptic compart-ment undergoes major rearrangement during NMJ maturation, including increases in the number, stabil-ity, density and clustering of AChRs; these changes are mediated by key molecular mechanisms.

The proteoglycan agrin is the most studied molecule involved in postsynaptic organization. Mice that lack agrin (Agrn−/− mice) show major defects in postsynaptic AChR organization: AChRs in these animals are uni-formly distributed on the surface of the muscle fibre and do not cluster after muscle innervation46. These mice also show presynaptic defects, such as poor arborization of motor axons46 and aimless overgrowth of nerves23. However, in these animals, the pre-patterned AChR clus-ters that form prior to innervation are not affected23,34,46. Interestingly, intramuscular injection of recombinant agrin induces the formation of AChR clusters in dener-vated muscles and the formation of ectopic AChR clus-ters in adult muscles47–49. Agrin can be synthesized and released by neurons, muscle fibres and Schwann cells (hereafter defined as SCs when referring to Schwann cells that are associated with axons or when the type is unclear), and has varying AChR-clustering efficacy depending on alternative splicing of the agrin gene50,51. Generally, neuronal forms of agrin are more potent at clustering AChRs11,52,53. However, AChR clustering can be enhanced by the active PSC-derived form of agrin — for instance, at the end of tadpole metamorphosis or during nerve regeneration at the amphibian NMJ11.

Agrin triggers the autophosphorylation of the mus-cle-specific tyrosine kinase receptor (MUSK)54. MUSK is enriched in the postsynaptic endplate, colocalizes with AChR55 and has a pivotal role in AChR clustering. Mice lacking MUSK (Musk−/− mice) show a more severe phenotype than do Agrn−/− mice: Musk−/− mice show nei-ther AChR pre-patterns nor nerve-induced clustering of AChRs, and their motor axons grow aimlessly over the muscle54. Agrin interacts with MUSK via low-density lipo-protein receptor-related protein 4 (LRP4), a transmem-brane protein that is enriched at the NMJ56–58. Mice that lack this lipoprotein die at birth and show major defects that are comparable to those seen in Musk−/− mice56. Indeed, LRP4 links agrin and MUSK activation, and is required for AChR clustering56–58. Furthermore, the asso-ciation of LRP4 with MUSK is sufficient to trigger MUSK

REVIEWS

704 | NOVEMBER 2014 | VOLUME 15 www.nature.com/reviews/neuro


autophosphorylation — even in the absence of agrin — and LRP4 is important for the pre-patterning of AChR clusters prior to the arrival of the nerve terminal57,58. As a tyrosine kinase receptor, activated MUSK influences numerous downstream signalling pathways (reviewed in REF. 59) and promotes the accumulation of synaptic mem-brane proteins such as rapsyn and neuregulin receptors. As well as being a self-associating protein, rapsyn can cluster AChRs, dystroglycan and other important NMJ proteins. Mice that are null for rapsyn show diffusely dis-tributed AChRs over the surface of the muscle fibre and have impaired neuromuscular transmission60 (reviewed in REF. 8).

WNT ligands are secreted molecules that regulate many aspects of embryonic development, including cell survival, cell death, cell proliferation and axon guidance, AChR aggregation and AChR clustering61,62, as well as the differentiation of the pre- and postsynaptic elements of the NMJ (reviewed in REFS 62,63). In zebrafish, WNT ligands interact with MUSK — before the nerve arrives at the NMJ — to induce the pre-patterning of AChRs on muscle fibres7,64. WNT proteins could subsequently par-ticipate in AChR clustering while nerve–muscle contacts form65,66. WNTs seem to be both positive and negative regulators of AChR clustering. In cultured mouse myo-tubes, WNT3, a WNT protein that is expressed by nerve terminals at the mouse NMJ, acts with agrin to increase AChR clustering65. By contrast, WNT3A, another iso-form of WNT, suppresses the expression of rapsyn and thus inhibits AChR aggregation at the surface of cultured myotubes66. Moreover, WNT can be released by motor neurons and by other cell types, including glial cells14,67. In Drosophila melanogaster, Wingless proteins (which are WNT orthologues) from glia at the NMJ facilitate the clus-tering of glutamatergic postsynaptic receptors14. However, although the roles of WNT proteins in postsynaptic matu-ration are gradually being revealed, their in vivo contribu-tions at the mammalian NMJ remain poorly understood owing to their diverse actions62,63 and cellular sources14,67.

The maturation of the NMJ must coincide with an increase in the synthesis of AChR and its associated proteins such as rapsyn and MUSK. Neuregulins — in particular, neuregulin 1 — that are expressed by motor neurons, muscle fibres and SCs68–70 may regulate the tran-scription of AChR-encoding genes and functional receptor expression70,71. Indeed, NMJs do not form properly in mice that lack neuregulin 1 or its receptors, ERBB2 and ERBB3 (REFS 16,17,72). Extracellular neuregulin 1 binds to these tyrosine kinase receptors of the ERBB family to promote the synthesis of AChR and its associated proteins68,71,73. However, the exact signalling pathway that is involved in promoting AChR synthesis and clustering remains elu-sive, and studies have questioned whether neuregulins and their receptors are necessary for NMJ formation15,74. For instance, the specific ablation of muscle ERBB2 and ERBB4 receptors in mice did not result in major defects15, whereas overexpression of a constitutively active form of the ERBB2 receptor in muscle fibres induced the aberrant formation of NMJs75. One study showed that motor-neuron-derived neuregulin 1 was not necessary for proper AChR density and clustering, and that muscle

neuregulin 1 was not necessary for postsynaptic matura-tion74. The ambiguity associated with the role of neureg-ulin 1 may be partly due to the multiple cellular targets and functions of neuregulin 1, notably in SC survival and maturation16,17,72,76. Indeed, SCs are absent in mice that lack neuregulin 1, ERBB2 or ERBB3, and NMJs in these mice do not form properly and are not maintained16,17,72. As PSCs express neuregulins and their receptors69, it has been proposed that the function of neuregulin 1 — in particular, its effect on postsynaptic maturation — may be mediated indirectly via PSCs8,15 (rather than by its direct effects on the muscle fibre). If this is the case, then both neuronal and PSC-derived neuregulins could be important for the maturation and maintenance of the NMJ, as well as for the regulation of the synthesis of AChR and its associated proteins. It remains unclear whether the NMJ defects observed in mice that are null for neuregulin 1, ERBB2 or ERBB3 are due to the absence of SCs or the lack of neu-regulin signalling, or both. Studies examining the effect of deleting PSC-derived neuregulin 1 or neuregulin 1 recep-tors at the surface of PSCs would help to resolve this issue. Nonetheless, it is apparent that a combination of molecules that are derived from various sources must act in a con-certed way to regulate and promote AChR clustering, as well as postsynaptic assembly and maturation (TABLE 1).

Although most of the mechanisms described above promote the synthesis and clustering of AChRs, other molecular signals promote the removal of improp-erly localized AChRs and limit the number of AChRs. Interestingly, this regulation is partly activity-dependent, as neuronal ACh release inhibits the expression of AChRs and promotes their endocytosis77,78. NMJs that lack choline acetyltransferase (ChAT) — which exhibit very low levels of ACh synthesis — are larger, more com-plex and have a broader distribution of AChR clusters, which suggests that ACh antagonizes AChR cluster-ing78–80. In fact, synaptic activity may trigger positive- or negative-feedback mechanisms depending on the state of postsynaptic AChRs. As the nerve–muscle contact is formed, nerve terminals release positive signals — such as agrin — that promote AChR clustering. However, if AChRs are not stabilized, then cholinergic neural activ-ity promotes their disruption and prevents further AChR synthesis and clustering. Interestingly, the phenotype observed in Agrn−/− mice — the uniform, non-clustered distribution of AChRs — is partially rescued in mice that lack both agrin and ChAT78. Such feedback mechanisms may also help to explain why AChRs are only found at synapses: other proteins, such as MUSK and rapsyn, ensure AChR stability and thus counteract the effect of cholinergic neural activity.

Similar to AChR clustering, the accumulation of agrin in the extracellular matrix (ECM) must be controlled to maintain proper receptor clustering. This may be achieved by matrix metalloproteinases (MMPs), which regulate the ECM by cleaving matrix proteins81. For instance, purified MMP3 can directly cleave agrin, lowering its levels in the ECM82–85. MMP3-null mutant mice show an increased density of AChRs, as well as agrin immunoreactivity82,85. MMP3 is secreted in its pro-form and requires cleavage to become active (reviewed in REF. 86). This activation

REVIEWS



might be activity-dependent, as denervated muscles exhibit high levels of agrin and presumably reduced levels of active MMP3 (REF. 85). Interestingly, MMP3 is found in the ECM around PSCs and is absent from active-zone areas82,83,85. Given the similar positioning of

PSC processes and MMPs at the edge of the synaptic cleft, it is tempting to propose that MMPs that are from, or acti-vated by, PSCs decrease perisynaptic levels of agrin, leav-ing it concentrated in the endplate area, where maximal AChR clustering is observed.

Table 1 | Molecular signalling for NMJ maturation

Molecules Sources Targets Roles Species Refs

Postsynaptic maturation

Agrin NT, MF and SC* MUSK–LRP4 complex MUSK activation, leading to ↑AChR clustering

Mouse, frog and D. melanogaster

11,23,50, 54

LRP4 MF and NT‡ MUSK MUSK activation, leading to ↑AChR clustering

Mouse 56–58,109, 110,112

WNT ligands MF, NT and SC§ MUSK (WNT9A, WNT11, WNT16) Agrin-dependent MUSK activation, leading to ↑AChR clustering

Mouse, zebrafish and D. melanogaster

61,64–66

MUSK (WNT3) Agrin-independent MUSK activation, leading to ↑AChR clustering

Target of WNT3A is not clear Suppression of rapsyn expression, leading to ↓AChR clustering

ACh NT AChRs ↓AChR synthesis and clustering Mouse 78–80,185

Neuregulin 1 NT and SC|| ERBB receptors ↑Synthesis and clustering of AChRs and associated proteins

Mouse 15,16,68, 71–74

TGFβ SC TGFBR1–TGFBR2 complex ↑Agrin expression Frog 12

Punt receptors ↑Agrin expression D. melanogaster 13

Synaptic laminins MF Integrins on MF (for laminin β2), dystroglycan on MF (for laminins α4 and α5)

↑AChR clustering Mouse 81,88,105

Presynaptic maturation

Synaptic laminins MF N−type (early) or P/Q‑type (late development) VGCCs on NT (for laminin β2); target for laminins α4 or α5 is not clear

• Regulation of active-zone proteins • Presynaptic organization

Mouse 6,88,89, 186

FGF7, FGF10 and FGF22

MF FGFR2B on NT ↑Vesicle clustering Mouse 93

Collagen α2, α3 and α6 chains (IV)

Not clear Not clear ↑Vesicle clustering Mouse 93

BDNF MF Pro-BDNF: p75NTR Retraction of NT Mouse 102,104

Mature BDNF: TRKB Nerve survival

GDNF MF and SC¶ Not clear Regulation of number of innervating axons Mouse 98,101,187

WNT ligands MF, NT and SC# Not clear Presynaptic branching and maturation D. melanogaster 113,114

LRP4 MF and NT Not clear • Presynaptic maturation (in vivo)• Synaptic vesicles and active-zone protein

clustering (shown in vitro)

Mouse 110,112

TGFβ SC TGFBR1–TGFBR2 complex Increase in nerve–muscle contacts and maturation

Frog and D. melanogaster

12,13

PSC maturation

Neuregulin 1 NT and SC ERBBs SC proliferation and survival Mouse 15–17

Synaptic laminins MF Not clear Laminins β2 and α4 or α5 control SC proliferation and migration

Mouse 19,20

ATP NT P2YRs SC activation Mouse 18

ACh, acetylcholine; AChR, ACh receptor; BDNF, brain-derived neurotrophic factor; D. melanogaster, Drosophila melanogaster; FGF, fibroblast growth factor; FGFR2B, fibroblast growth factor receptor 2B; GDNF, glial‑cell‑derived neurotrophic factor; LRP4, low‑density lipoprotein receptor‑related protein 4; MF, muscle fibre; MUSK, muscle‑specific tyrosine kinase receptor; NMJ, neuromuscular junction; NT, nerve terminal; P2YR, purinergic type 2Y receptor; p75NTR, p75 neurotrophin receptor; PSC, perisynaptic Schwann cell; SC, Schwann cell; TGFβ, transforming growth factor-β; TGFBR, transforming growth factor-β receptor; TRKB, tropomyosin‑related kinase B; VGCC, voltage‑gated calcium channel. *Neuronal isoform is more potent, muscle isoform is dispensable and SC isoform enhances AChR aggregation. ‡Expression of neuronal LRP4 has not been determined (see main text for details). §At the D. melanogaster NMJ, SC-associated WNT (Wingless) signalling regulates glutamate receptor clustering. ||Not yet clear. ¶Overexpression of GDNF by SCs did not cause hyperinnervation. #The SC-derived WNT-induced contribution to presynaptic regulation remains unknown.

REVIEWS



Synapse eliminationA reduction in the number of synaptic contacts that results from activity-dependent synaptic competition.

Presynaptic maturation. Presynaptic nerve terminals also undergo major changes during NMJ formation. Specifically, there is an increase in the number of nerve ramifications and active zones, and in the levels of asso-ciated proteins; there are changes in the distribution and types of calcium channels expressed; and there is a decrease in the number of nerve terminals in contact with a single muscle fibre. Unlike postsynaptic matura-tion, fewer molecular signals that control presynaptic differentiation have been identified, and the molecular mechanisms that underlie presynaptic maturation are less well understood.

Laminins are glycoproteins that are considered to be important regulators of presynaptic terminal maturation. They are heterotrimeric (comprising α, β and γ subunits) muscle-derived glycoproteins and are included in the basal lamina of the ECM. The absence of laminins results in mild to severe defects in NMJ development6,19,87,88. For example, NMJs from mice that lack the laminin β2 chain (which is present in all synap-tic laminins) have fewer active zones, synaptic vesicles that do not cluster and reduced synaptic activity. In addition, the processes of SCs in these animals invade the synaptic cleft87,88. Laminin β2 binds to presynaptic N-type voltage-gated calcium channels during early stages of development and to P-type and Q-type chan-nels at later stages81,89 (different channels are involved in transmitter release at these respective stages of devel-opment)89,90. The interaction of laminin β2 with cal-cium channels regulates active-zone proteins such as bassoon and the distribution of calcium channels89,91. Other synaptic laminins, such as laminin α4 and laminin α5, are required for NMJ maturation and for the proper alignment of pre- and postsynaptic elements (reviewed in REF. 81).

Certain fibroblast growth factors (FGFs) — namely, FGF7, FGF10 and FGF22 — promote NMJ maturation and vesicle clustering in motor neurons in vitro92,93 and in vivo during embryonic and early postnatal devel-opment, as revealed using mice lacking FGF receptor type 2B (FGFR2B; Fgfr2b−/− mice)93. Muscle fibres are the main sources of FGFs, which signal at the NMJ mainly through presynaptic FGFR2B93,94. Another mol-ecule involved in NMJ maturation is type IV collagen, which is found primarily in the basal lamina93,95. Vesicle clustering is promoted by collagen IV α2 chains during embryonic stages, and by collagen IV α3 and α6 chains during postnatal development93. Interestingly, FGFs (FGF7, FGF10 and FGF22), laminin β2 and collagen IV (α2, α3 and α6 chains) were all found to be important for presynaptic maturation, and the three classes of mol-ecules were shown to act sequentially from embryonic to postnatal stages93. At least for FGF7, FGF10, FGF22 and collagen IV (α3 and α6 chains), this idea is con-sistent with the levels of expression of mRNA encod-ing these proteins at different developmental stages93. Type XIII collagen may also be important for pre- and postsynaptic maturation, as NMJs from mice that lack collagen XIII have fewer active zones, abnormal AChR clustering and improper presynaptic–postsynaptic alignment, all of which correlate with synaptic defects96.

Finally, the extracellular domain of the signal regulatory protein-α (SIRPα), which is a member of the immuno-globulin family, promotes synaptic-vesicle clustering in cultured chick spinal cord motor neurons97.

Glial-cell-derived neurotrophic factor (GDNF) and brain-derived neurotrophic factor (BDNF) also regulate presynaptic maturation. GDNF is expressed by muscle fibres and SCs, and promotes motor neuron survival98,99. Injection of GDNF, or the genetic overexpression of its muscle isoform, leads to an increase in the number of nerve terminals innervating the same NMJ and delays synapse elimination100,101. SCs also express GDNF, but the role of SC-derived GDNF in NMJ maturation remains unclear101. Like GDNF, the application of BDNF to the muscles of neonatal mice delays synapse elimina-tion102. BDNF is initially synthesized as its precursor form (pro-BDNF), which is then cleaved to generate the mature form (mBDNF)103. Through different receptor interactions, pro-BDNF favours synapse elimination, whereas mBDNF promotes the maintenance of syn-apses102,104. Moreover, the conversion from pro-BDNF to mBDNF may be regulated by synaptic MMP3 and MMP9 (REFS 102,104). As MMP3 localizes in the ECM that surrounds PSCs85, it is again appealing to propose that PSC-derived MMPs might convert pro-BDNF to mBDNF and thus influence presynaptic maturation.

Coordinated tri-directional maturation of the NMJ. In order to build synapses that are adapted and functionally tuned, the maturation of the NMJ must be well coor-dinated. Many molecules and pathways, from all three synaptic elements at the NMJ, are involved in aspects of this process (FIG. 1; TABLE 1).

Muscle-derived laminins that contain α4, α5 or β2 chains have an autocrine role in postsynaptic matu-ration105. Laminins containing the β2 chain bind to inte-grin receptors containing the β1 subunit in the ECM to increase AChR clustering, and laminins with an α4 or α5 chain bind to dystroglycan, which stabilizes AChRs in the postsynaptic membrane and participates in the recruitment of acetylcholinesterase to the NMJ106–108 (reviewed in REF. 81). Thus, the anterograde and retro-grade effects of synaptic laminins may be responsible for proper pre- and postsynaptic alignment. Laminins also control glial maturation: synaptic laminin 521 (which comprises α5, β2 and γ1 subunits) actively inhibits PSCs from extending their processes into the synaptic cleft19. In mice that lack the laminin β2 chain, the inva-sion of the NMJ synaptic cleft by PSC processes could explain many of the abnormalities in NMJ function, as PSCs could block synaptic transmission and neuron– muscle interactions88. Therefore, laminins coordinate the maturation of the NMJ and ensure the correct relative positioning of each synaptic element19.

Like laminins, LRP4 acts bidirectionally to coordinate pre- and postsynaptic development. LRP4 acts as an agrin co-receptor to stimulate AChR clustering57,58,109, but it is also necessary for presynaptic maturation (which is reflected by the clustering of synaptic vesicles and active-zone proteins)110. Muscle-derived LRP4 may act as a ret-rograde signal to regulate presynaptic maturation in a

REVIEWS



MUSK-independent manner110. By contrast, MUSK may indirectly participate in presynaptic differentiation by clustering LRP4 (REFS 59,110,111), thus concentrating the bidirectional signalling of LRP4. Although the expression of LRP4 by muscle fibres has been established56,57, a recent study suggests that neuronal LRP4 may also contribute to NMJ maturation112. More-severe NMJ deficits were observed in mutant mice that lacked both muscle- and motor-neuron-derived LRP4 than in mice that only lacked muscle-derived LRP4 (REF. 112). However, direct evidence of LRP4 protein or RNA expression specifically in motor neurons remains undetermined112.

WNT ligands may also have effects beyond AChR clustering. For instance, disrupting Wingless protein activity in D. melanogaster results in both pre- and post-synaptic NMJ defects, which suggests that, at least in flies, Wingless proteins serve as both anterograde and retrograde signals to promote NMJ formation113,114. BDNF and GDNF could also have multiple roles at developing NMJs, as muscle fibres and SCs express receptors for both neurotrophic factors115–117. In neuron–muscle cultures, GDNF increases the membrane levels of AChRs in the muscle fibre without altering AChR synthesis118, whereas in cultured myotubes BDNF can

Figure 1 | Synapse formation at the NMJ: generating a functional communication unit. Parallel signalling pathways at the neuromuscular junction (NMJ) are involved in the coordinated maturation of the presynaptic terminal (green), the postsynaptic muscle fibre (red) and the perisynaptic Schwann cells (PSCs; grey); dotted lines indicate pathways that are still being debated. Agrin, which is released by the nerve terminal, muscle fibre and surrounding PSCs, acts on the LRP4–MUSK complex, which comprises low‑density lipoprotein receptor‑related protein 4 (LRP4) and muscle‑specific tyrosine kinase receptor (MUSK). The phosphorylation of MUSK leads to rapsyn-mediated clustering of ionotropic acetylcholine receptors (AChRs) and postsynaptic maturation. AChR clustering can also be enhanced by WNT ligands (associated with MUSK in the figure), whereas release of ACh inhibits AChR clustering. LRP4 is mainly of postsynaptic origin, although neuronal LRP4 may also have a role (see main text for details). LRP4 acts as a co-receptor for agrin and stimulates AChR clustering as well as presynaptic maturation by clustering synaptic vesicles and active-zone proteins (not shown). Neuregulin 1 (NRG1) can be released by the nerve terminal and/or surrounding PSCs, and binds to PSC‑expressed or postsynaptic ERBB receptors (ERBB2 or ERBB3). The binding of NRG1 to ERBB receptors on PSCs promotes PSC survival and maturation. Although the exact signalling pathway is controversial, the activation of postsynaptic ERBB receptors by NRG1 may increase levels of postsynaptic proteins such as rapsyn, MUSK and AChR, and could have a role in AChR clustering, which overall leads to postsynaptic maturation. Synaptic laminins, such as laminins β2, α4 and α5, are released by the muscle fibre. They form heterotrimeric glycoproteins that are included in the basal lamina and are important for proper pre- and postsynaptic alignment and maturation, as well as PSC maturation (not shown). Laminin β2 binds to presynaptic calcium channels and regulates active-zone proteins (not shown). Postsynaptically, laminins interact with integrin β1, which increases AChR clustering. Fibroblast growth factors (specifically FGF7, FGF10 and FGF22) are released by the muscle fibre and activate mainly presynaptic type 2B FGF receptors (FGFR2B); thus, they are important for vesicle clustering and presynaptic maturation, as are type IV collagen α chains (α2, α3 and α6). PSC-derived transforming growth factor-β (TGFβ) induces presynaptic maturation and postsynaptic differentiation by upregulating the expression of agrin. Synaptically released ATP is detected by PSC-expressed purinergic type 2Y receptors (P2YRs) and triggers increases in intra-PSC Ca2+ concentrations. PSCs also express muscarinic AChRs (mAChRs), and their activation by the local application of ACh triggers increases in intra-PSC Ca2+ concentrations. However, mAChRs are not activated by endogenous ACh release. Matrix metalloproteinases (MMPs) in the extracellular matrix (ECM) that surrounds PSCs regulate the composition of the ECM and cleave matrix proteins such as agrin, triggering its removal from the ECM.

RapsynAChR

AChATP

Agrin

ERBB

TGFβ

Ca2+ channel

β1 integrin

β2 laminin

Type IV collagen α

FGFR2B

ECM

Nerve terminalPSC

Muscle bre

Presynaptic maturation

Survival and maturation

Postsynaptic maturation

NRG1

WNT

LRP4

MUSK

FGF

↑[Ca2+]

MMP

mAChR

P2YR

Upregulationof agrin

REVIEWS



GangliosideMember of a family of oligo-glycosylceramide plasma-membrane lipids that was originally discovered after its isolation from ganglion cells and which is predominantly found in the nervous system. Antibodies against some disialosyl epitopes of gangliosides can be used to specifically ablate perisynaptic Schwann cells at the mammalian neuromuscular junction.

inhibit agrin-induced AChR clustering115 and maintain AChR clusters117. Moreover, the regulation of PSC activ-ity by neurotrophic factors at the adult mouse NMJ119 indicates that such molecules may have broader effects on NMJ synaptogenesis.

As highlighted above, PSCs participate in many important steps of synapse maturation. SCs rely on pre- and postsynaptic elements for survival and maturation (BOX 1; TABLE 1), but they also contribute to pre- and postsynaptic maturation and NMJ maintenance. For instance, SCs are absent in mice that lack neuregulin 1, ERBB2 or ERBB3. Although contact formation seems to occur normally in these mice, nerve terminals sub-sequently retract and motor neurons degenerate shortly afterwards16,72. This strongly suggests that SCs are not important for the initial process of synapse formation but are essential for the survival of motor neurons and for the maintenance of NMJs during synapse maturation. SCs in general are important for neuronal survival120,121, and PSCs may have similar roles at the NMJ. One study examined the contribution of PSCs at the NMJ using

complement-mediated cell lysis to specifically ablate PSCs at the amphibian NMJ in vivo10. Without PSCs, the NMJs were smaller, had fewer branches and underwent a long-term loss of synapses10. Similar observations were made at the mouse NMJ following selective PSC ablation with a ganglioside-specific autoantibody122,123.

Interestingly, transforming growth factor-β1 (TGFβ1) has been proposed to mediate some of the synaptogenic effects of PSCs on the NMJ. TGFβ1 is expressed by PSCs, found in SC-conditioned media and promotes synaptogenesis in nerve–muscle co-cultures12. Indeed, in cultured spinal neurons treated with SC- conditioned media, TGFβ1 increased the size and num-ber of AChR clusters at nerve–muscle contact sites, probably by upregulating agrin expression12. In vivo, glial-cell-derived TGFβ was also shown to be a synapto-genic factor and modulated synapse development at the D. melanogaster NMJ. Glial TGFβ activates muscular TGFβ receptors and thus increases retrograde muscle signalling and facilitates synaptic growth13. Moreover, recent data from our laboratory suggest that PSCs may

Box 1 | Maturation of PSCs

Perisynaptic Schwann cells (PSCs) are generated from stem cells that originate from the neural crest and follow motor

axons into the muscles, where they reach their final location at the neuromuscular junction (NMJ) (reviewed in REFS 5,162). Immature Schwann cells begin to take on a PSC phenotype once they reach their location at the NMJ. As the

number of PSCs at an NMJ increases18,150,151, their

shapes become increasingly complex, their

motility decreases and their territories become

segregated (see the panels in the figure

illustrating the changes in the number18,149,151,

shape149 and territory149 of PSCs). These changes

have been attributed to the spatial constraints

that dictate PSC territories, suggesting that PSCs

actively compete for available space149.

Interestingly, this period of PSC remodelling

coincides with the peak of postnatal maturation

of the NMJ; in mice, this occurs during the first

postnatal week. During this period, PSCs show a

high level of phagocytic activity146 when pruning

connections during synapse maturation21,22, and

they express receptors to detect and decode

neurotransmission18.

Interestingly, the PSC receptors that detect

neurotransmission differ between developing and

mature NMJs (see the figure panel labelled

‘excitability’)18,154. Whereas PSCs at immature

NMJs rely mainly on purinergic signalling18, PSCs

at mature NMJs use purinergic receptors

(including purinergic type 2Y receptors (P2YRs)) and muscarinic metabotropic acetylcholine

receptors (mAChRs) to detect released neurotransmitters154. This differentiation may be

important for regulating the expression of a

number of genes, particularly those under the

control of mAChR activation163, such as the gene

encoding glial fibrillary acidic protein (GFAP). One

intriguing possibility is that the receptor types

involved may be adapted to the state of NMJ

plasticity, such that low mAChR activity is permissive for growth and remodelling, whereas

higher mAChR activation promotes stability and the modulation of synaptic activity and efficacy.

Nu

mb

er

Muscle bre

Terminal 1

Terminal 2

PSC

ndplate

Sh

ap

e

Fried‑egg shape

ndplate

PSC1 PSC2

PSC3

Te

rrit

ory

Intermingled territory

Muscle bre

PSCNerveterminal

Gene regulation

Rapsyn

AChR

AChR cluster

ATP

ACh

↑ Ca2 ] P2YR

mAChR

Gene regulation

↑ Ca2 ]

Ex

cit

ab

ilit

y

Complex and branched shape

Segregated and delineated territories

Mature NMJDeveloping NMJ

REVIEWS



Retraction bulbEnlarged distal part of the axon undergoing retraction (that is, axosomal shedding); it is commonly observed during synapse elimination.

either permit or restrict NMJ expansion and that these effects depend on the activation of PSCs by released neurotransmitters (ACh or ATP) and on the maturity of the synapse18 (FIG. 1).

Activity-dependent NMJ maturationAlthough the NMJ is functional upon its formation, it undergoes maturation for a prolonged period. Like all synapses in the nervous system, it depends on synaptic activity to develop an optimal organization and proper synaptic efficacy to accomplish its function. Activity shapes both the postsynaptic apparatus8,78,79 and presyn-aptic organization, as well as synaptic function. One of the most fascinating activity-induced changes at the NMJ is the dramatic decrease in the number of synaptic

connections. In mature NMJs, a single nerve terminal contacts the postsynaptic cell; however, during early stages of development, each muscle fibre is innervated by several nerve terminals that branch from different motor neu-rons44,45. As a result, competition ensues between nerve terminals for the same postsynaptic endplate area. This competition is activity-dependent and leads to the elimi-nation of supernumerary inputs; it is regulated by pre- and postsynaptic elements and by PSCs associated with the synapse. A similar activity-dependent competition occurs between regenerated axons during the reinnervation of muscle fibres following nerve injury (BOX 2).

Unlike the pathways that regulate pre- and postsyn-aptic maturation, little is known about the molecular mechanisms that regulate activity-dependent synaptic competition. Moreover, the relationship between the activity-dependent processes and the molecular path-ways that drive synapse formation and maturation remains ill defined. We discuss below synaptic competi-tion and elimination, with a particular focus on the pos-sible underlying molecular mechanisms and the role of glial cells during these processes.

Synaptic competition and elimination. Between the late embryonic period and the early postnatal period, nerve terminals at the same NMJ compete with one another for the sole innervation of the muscle fibre4,45,124,125. This pro-cess is important for the development of proper synaptic connections and synaptic function43,44,126,127. In vivo time-lapse imaging of synaptic elimination at dually inner-vated NMJs revealed that one axon progressively vacates a synaptic area, which is then gradually taken over by a competing nerve input125,128. Typically, a ‘losing’ terminal ultimately retracts in the form of a ‘retraction bulb’, which is eliminated by PSCs21,125. This competition process is highly dynamic, and competing terminals continuously vie for territories. Indeed, the outcome of the compe-tition is not preordained, and the territories of nerve terminals can even shift back and forth125, such that a terminal that initially occupies a smaller area could take over the territory and ultimately ‘win’ the competition125. Interestingly, a recent study showed that terminals that were considered to be losing the competition (that is, smaller terminals) — and even retraction bulbs at a dis-tance from the endplate — were able to reverse their fate and grow back to occupy the NMJ if their competitor was ablated129.

The difficulty in predicting a winning nerve termi-nal solely on the basis of morphological and structural observations suggests the mechanisms involved in syn-aptic competition and elimination are complex. It has been suggested that the outcome of synaptic competition and elimination depends on the efficacy and structure of nerve terminals — properties that are dependent on neuronal activity130–132. The general idea is that, at early stages of synaptic competition (for example, during the first postnatal days in mice), terminals that compete at the same NMJ could have comparable synaptic effica-cies and presynaptic areas that occupy equally sized postsynaptic territories. Then, one input is gradually strengthened and occupies a larger area at the expense

Box 2 | NMJ formation following injury and in disease

The peripheral nervous system has a remarkable ability to regenerate164,165. Following

nerve damage, the distal sections of injured axons remain functional for a given period

before Wallerian degeneration occurs. This time period is known to directly depend on

the length of the degenerating nerve stump166,167. Wallerian degeneration involves many

events: changes to the blood–nerve barrier, such as increased permeability; the

degeneration of the distal axon; the dedifferentiation and proliferation of Schwann

cells, which start invading the synaptic cleft168,169; the recruitment of macrophages; and

the reorganization of the extracellular matrix170. The combined goal of all of these

events is to generate an innervation-permissive environment so that regenerated axons

can reinnervate the muscle endplate once Wallerian degeneration is complete.

During reinnervation, axons continue to grow beyond the boundary of a denervated

neuromuscular junction (NMJ; which is delineated by the endplate area and the

clustered acetylcholine receptors (AChRs)), where they travel along peripheral Schwann cell bridges that have been formed during denervation165. These ‘escaped

fibres’ can reach another synaptic site, leading — along with the axonal reinnervation

— to the polyinnervation of NMJs171,172.

Remarkably, reinnervation shares several features with synapse formation during development. First, in both cases, NMJs that are polyinnervated undergo a period of

synaptic competition, which leads to monoinnervation171–173. Second, as with

elimination in the developing NMJ, elimination during reinnervation depends on and is

shaped by synaptic activity137,174. Therefore, there is evidence to suggest that the

processes that occur during NMJ formation are recapitulated following nerve injury.

This would imply that during reinnervation perisynaptic Schwann cells (PSCs) might

adapt their sensibility in response to neurotransmission until it resembles that observed

in immature PSCs. This suggests that the relative contributions of muscarinic and

purinergic signalling in PSCs can also be adapted.

Although these NMJ-forming processes can be repeated, severe diseases of the NMJ

can arise if a single step of NMJ formation is compromised. For example, the production

of autoantibodies targeting specific proteins that are present in the endplate leads to

neuromuscular diseases such as myasthenia gravis, which is characterized by muscle

weakness and fatigue. Autoantibodies that are involved in the more prevalent

neuromuscular diseases include those that target AChRs175, muscle-specific tyrosine

kinase receptor (MUSK)111,176–178 and low-density lipoprotein receptor-related protein 4 (LRP4)179,180 (reviewed in REF. 181). Furthermore, mutations in the genes encoding

proteins that are essential for NMJ formation, such as MUSK59 or proteins associated

with the basal lamina, lead to congenital neuromuscular diseases182. Recently, motor-endplate disease, a severe genetic disorder in mice that is characterized by

progressive muscular weakness and that leads to paralysis and premature death, has

been attributed to a mutation in the gene encoding the voltage-gated sodium channel

Nav1.6. Interestingly, this mutation leads to the apoptosis of PSCs during synapse

formation183,184.

Thus, a better understanding of NMJ formation and maturation will help to elucidate

both the regenerative processes that occur after nerve injury and how altered

maturation leads to major defects. Moreover, considering the NMJ as a functional unit

composed of a presynaptic terminal, a muscle fibre and associated PSCs may lead to

the identification of additional therapeutic targets to treat neuromuscular diseases.

REVIEWS



Spike timingPattern of temporal correlation between presynaptic and postsynaptic activities. Synchronous and asynchronous patterns of activity are observed during early and late phases of synapse elimination, respectively.

Asynchronous activityUncorrelated timing of synaptic inputs onto the muscle fibre, leading to an out-of-phase activation.

of the other nerve terminals44,133. At dually innervated NMJs, a disparity in the synaptic efficacies of compet-ing inputs gradually builds up, resulting in a clear dis-tinction between a strong input (with a high release of neurotransmitter) and a weak input (with a low release of neurotransmitter)18,130,132 (FIG. 2).

Although initially a controversial idea134–136, there is evidence to suggest that synaptic inputs that are more active are favoured during synaptic competition131,137–141. For instance, Buffelli et al.131 deleted ChAT in a subset of motor axons in mice to cause a robust neuronal-subset-specific reduction in neurotransmission and hence a disparity in synaptic efficacy between affected and unaf-fected subsets. This manipulation resulted in maintenance of the more-active inputs and elimination of the others131. In addition to relative synaptic efficacy, the pattern of activity of competing terminals (spike timing) is an impor-tant determinant of synapse elimination. During the early phase of synaptic competition (postnatal day 0 (P0) to P4),

motor-unit activity from competing terminals is synchro-nous, but these inputs become gradually asynchronous during the period when synapse elimination is prominent (P4–P10). Motor units reach complete asynchrony at the end of the synaptic competition and elimination process (at ~P14 in mice)139,142. Interestingly, inducing asynchrony between spikes from competing axons promotes synapse elimination, whereas imposed synchrony extends the period of synaptic competition137.

Asynchrony — that is, distinct firing patterns from different presynaptic terminals — could allow the mus-cle fibre and/or PSCs to detect and distinguish between independent inputs and then actively participate in the strengthening or weakening of individual terminals. For example, focal blockade of AChRs in a portion of an endplate area caused presynaptic retraction from this specific area, whereas blockade of an entire endplate area did not induce presynaptic loss138. This suggests that the postsynaptic element must distinguish between

Figure 2 | Synaptic competition and elimination at the NMJ. Morphological and functional changes during postnatal maturation are associated with synaptic competition and elimination. a | A dually innervated neuromuscular junction (NMJ) with two nerve terminals (green and blue) competing for the same endplate (represented by the darker red on the pale red muscle fibre) is represented schematically. Competing terminals start with similar territories (left panel; territories could be intermingled, but this is not shown here). Gradually, territories segregate and one terminal (blue) occupies a larger area (middle panel), leading to the elimination of the losing axon (right panel). b | Competing terminals could have a similar synaptic efficacy (left panel), as defined by the small differences in quantal content ‘m’, the similar rate of failed and evoked endplate potentials (EPPs), and the similar paired pulse facilitation (PPF) profiles of each of the terminals (left graph). One input (blue) gradually becomes stronger (middle graph), with a larger quantal content, lower failure rate and lower PPF (which is defined as the increase in the amplitude of the second EPP compared to the first one) and a larger number of evoked EPPs with higher amplitudes. (Experiments supporting these observations were performed in a condition of low probability of release, with low Ca2+ and high Mg2+ concentrations)18,130,132. Once the losing terminal (green) retracts (right graph), it no longer induces synaptic activity at this given NMJ (100% failure). c | Patterns of neuronal firing change from synchronous motor‑unit activity (left panel) to asynchronous activity (middle and right panels), which leads to synapse elimination. The functional changes depicted in parts b and c are indirectly associated with the morphological changes presented in part a.

Retraction bulb

Muscle bre

Terminal 1 Terminal 2Territory occupancya

Synaptic efficacyb

Even

t fre

quen

cy (%

)

EPP amplitude (mV)

Failed

Evoked

Even

t fre

quen

cy (%

)

EPP amplitude (mV)

Failed

Evoked

Even

t fre

quen

cy (%

)

EPP amplitude (mV)

Failed

Evoked

m1 m2 PPF

mVms

m1 < m2 Weak Strong

Spike timing patternc

Endplate

Synchronous Asynchronous

REVIEWS



inputs to promote survival or elimination. More-active nerve terminals are also more likely to fire out of syn-chrony with their ‘quieter’ competitors; therefore, a number of spikes will occur in the most active input when its competitors are silent137. Thus, both the level and the pattern of activity shape synaptic competition and elimination.

Although a direct link between synaptic strength, structure and elimination has yet to be elucidated, there is evidence to support the notion that these three factors are interdependent44,129,131. For example, a slight decrease in synaptic activity could weaken the interaction between a terminal and a muscle fibre, resulting in a vacated ter-ritory as the less active terminal is remodelled. This terri-tory could then be occupied by the competing terminal, and this new occupation may increase the feedback from the muscle fibre and PSCs to reinforce this terminal and result in improved efficacy. An experiment in which both synaptic activity and the territory occupancy of compet-ing terminals could be monitored and modulated would provide valuable information on the relationship between synaptic organization and activity during synapse compe-tition. In parallel with the effects of synaptic activity and territory occupancy, competing terminals may constantly battle against each other for survival factors such as GDNF and BDNF101,102,104. Indeed, pro-BDNF and mBDNF serve as punishment and reward signals, respectively, to either eliminate or maintain competing nerve terminals, leading to the maintenance of more-active inputs102,104. Thus, this would ultimately favour a nerve terminal that has a larger synaptic strength and a better pattern of synaptic activity, and that is in a more advantageous position to benefit from survival factors.

The process of synaptic competition requires a con-stant remodelling of the NMJ such that the molecular mechanisms that are involved in synapse maturation may also be dynamically regulated during synaptic compe-tition. The elimination of nerve terminals must involve activity-dependent molecular mechanisms to weaken some nerve–muscle interactions while stabilizing oth-ers. Hence, one could hypothesize that the mechanisms involved in AChR clustering and stabilization (includ-ing MUSK signalling and receptor anchoring by rapsyn) are regulated in an activity-dependent manner. As some terminals are eliminated, the postsynaptic territory can be either re-occupied by neighbouring inputs or left vacated22,125. If left unoccupied, the postsynaptic appara-tus will be remodelled: the improperly localized AChRs will be dispersed and removed125,143. Consistent with this possibility, during postnatal development the end-plate area changes from an ‘en plaque’ oval structure to a branched ‘pretzel-like’ shape owing to the loss of AChRs from some regions143. As ACh regulates the expres-sion and clustering of AChRs, this transition could be activity-dependent77–80. Indeed, this physical transforma-tion temporally corresponds with synaptic competition and elimination events, which themselves are activity-dependent8,143. Moreover, activity-dependent changes may be regulated by many molecules, such as laminins, as the transition into a pretzel-like shape is delayed at NMJs in mice that lack laminin α5 (REF. 105).

Glial roles in synaptic competition and elimination. As axons gradually retract, they leave behind debris-contain-ing synaptic organelles known as axosomes21,144 (FIG. 3). These are formed by the engulfment of axonal endings by PSCs and are subsequently entirely contained within these cells21. Axosomes have been observed following the elimination of terminal branches — leading to segrega-tion of the territories of competing terminals21 — and during the retraction of nerves as retraction bulbs, which leads to synapse elimination21 (FIG. 3a). Consistent with the well-known role of glial cells in debris clearance145, the retraction of terminals is associated with high lysoso-mal activity, which suggests that axosomes are processed through the lysosomal pathway of SCs22,146. Interestingly, axonal debris was cleared more slowly in a mouse model of a lysosomal storage disease146. At the fly NMJ, the phagocytosis of presynaptic debris results from coor-dinated efforts of glial and muscle cells147,148. These cells phagocytose presynaptic debris, including debris from weak and unstable synaptic boutons. This is an activity-dependent phenomenon, as the accumulation of debris increases following high-frequency stimulation of motor neurons148.

With regard to synapse reorganization and elimina-tion, a key question is how the nerve terminals to be maintained — or pruned and eliminated — are identified. Two recent studies implicate PSCs in synapse elimination. First, Smith et al.22 showed that during development, PSCs arrive at the NMJ and cover all terminals with no appar-ent preference. Early in the process of synapse elimina-tion (that is, during embryonic and early postnatal stages), PSCs extend their processes between the nerve terminals and muscle fibres, obstructing AChRs and synaptic com-munication. It has been proposed that PSCs compete with nerve terminals to be in close proximity to AChR-covered surfaces in order to make even more contact with AChRs than the terminal boutons. PSCs then phagocy-tose and remove synaptic branches from different termi-nals, resulting in the random elimination of terminals. The eliminated terminals leave behind vacant territory that can be occupied by non-eliminated neighbouring inputs22 (FIG. 3b). Interestingly, given the tight relationship between AChRs and glial coverage, it has been proposed that PSCs may also sculpt the postsynaptic endplate22, perhaps via an increase in MMP3 activity, which would cause the cleavage of agrin and thus the elimination of unstable AChRs83,85. Such mechanisms could contribute to the transition of the endplate into a pretzel-like shape, and would be consistent with the change of PSCs from a ‘fried-egg’ shape to a branched morphology during post-natal development149 (BOX 1). However, these hypotheses remain to be tested.

Smith et al.22 propose that PSCs do not select a win-ning nerve terminal per se but instead randomly elimi-nate branches, eventually leading to a sole survivor. This study provides extensive and important information on the ultrastructural organization of polyinnervated NMJs. It also highlights the complex structural interac-tion between competing nerve terminals, the postsyn-aptic apparatus and the surrounding PSCs. However, it does not fully explain the role of neuronal activity in

REVIEWS



PSC-receptor segregationSpatial grouping and separation of receptors on the surface of perisynaptic Schwann cell (PSC) processes. Presumably, this grouping enables the PSC to detect neurotransmitter release from each competing terminal at dually innervated neuromuscular junctions. This segregation was described for purinergic type 2Y receptors, which mediate intra-PSC calcium activity during synaptic competition.

this scenario. Our group proposes another mechanism by which PSCs might be involved in synapse competi-tion and elimination18. As weaker terminals are more likely to be eliminated, PSCs may be able to differenti-ate between weaker and stronger terminals, and could therefore actively participate in the targeted elimination of ‘losing’ synapses. PSCs must possess certain qualities to actively decide which synapses to eliminate. As exem-plified by the fact that most NMJs on the soleus muscles are covered by a single PSC during this period18,150,151, a single PSC must be able to discriminate between com-peting inputs and accurately compare their relative syn-aptic strengths (as relative synaptic strength is a major determinant of synaptic competition). This condition was tested at dually innervated NMJs using a nerve–muscle preparation in which the activity of individual terminals could be independently evoked and recorded while observing the increases in intra-PSC Ca2+ levels (which reflect PSC activity) that were evoked by syn-aptic activity152–154. We showed that a single PSC at a dually innervated NMJ detects neurotransmitter that is released from each nerve terminal18 (FIG. 4). Moreover, Ca2+ responses that were induced within the same PSC by neurotransmission from different competing termi-nals were proportional to the relative synaptic efficacy of each terminal (FIG. 4).

This tight relationship between terminal synaptic efficacy and intra-PSC Ca2+ responses does not only cor-respond to the amounts of neurotransmitter released, but also depends on intrinsic PSC properties, including PSC-receptor segregation18. Indeed, a single PSC seems to organize its receptors in functional groups to face each competing terminal18, an arrangement that could enable a more efficient detection of released neurotransmitter and a more accurate differentiation of competing inputs. When more than one PSC is present at a dually inner-vated NMJ, each PSC differentiates the weak input from the strong input using similar mechanisms18. Thus, PSCs integrate synaptic properties by constantly monitoring the strength of competing terminals. This also implies that PSCs must provide specific feedback that either pro-motes or represses synapse elimination. Although such a feedback mechanism has not yet been elucidated, PSCs could target elimination machinery specifically towards weaker inputs, or — as PSCs can regulate synaptic activ-ity at mature NMJs (see below)153,155 — they could dif-ferentially modulate the efficacy of competing terminals, preferentially strengthening one terminal over the other.

The mechanisms proposed by Smith et al.22 and by our group18 are not mutually exclusive and could be complementary during synaptic competition. Indeed, at late embryonic and early postnatal stages, many motor

Figure 3 | PSCs eliminate supernumerary connections. a | A schematic representation of the losing nerve terminal shedding axosomes21,144 is shown. Step 1: during synaptic competition, competing nerve terminals (in green and blue) may occupy similar territories of the muscle fibre endplate (red) at the neuromuscular junction and are covered by a perisynaptic Schwann cell (PSC; translucent grey). Step 2: PSCs are thought to engulf the endings of nerve terminals, leaving organelles containing synaptic debris (axosomes) that are processed by the PSC lysosomal pathway (not shown). This decreases the size of the territory occupied by the losing terminal (terminal 1; green). This territory can be gradually occupied by its competitor (terminal 2; blue). Step 3: PSCs continue to phagocytose the losing terminal until the nerve retracts into a retraction bulb, which is ensheathed by Schwann cell processes. Step 4: following the elimination of the losing terminal, a single terminal (terminal 2; blue) innervates the muscle fibre. b | Mechanism of synapse elimination by PSCs22 is shown. Step 1: a schematic representation shows a PSC (translucent grey) covering two nerve terminals (terminal 1, green; terminal 2, blue), which are competing to innervate the same muscle fibre (red). Steps 2 and 3: the PSC seems to compete with nerve terminals for acetylcholine receptors (AChRs; red band) by extending its processes between the nerve terminals and the muscle fibre, partially blocking synaptic neurotransmission. Step 4: the distribution of AChRs can be remodelled in places that no longer receive neurotransmitters. In parallel, the PSC phagocytoses nerve endings (in this case, the green terminal) that are in contact with the muscle fibre, on a stochastic basis22. Step 5: the vacated area can be occupied by a nearby terminal (blue).

Muscle bre

Terminal 1 Terminal 2

Retraction bulb

a Axosome shedding

PSCAxosome

ndplate

1 2 3 4

b Synapse elimination by PSCs

Muscle breAChRs

Terminal 2Terminal 1

PSC

1 2 3 4 5PSC phagocytoses terminal

Redistribution of AChRs

REVIEWS



Muscle-twitch tensionTension elicited by a muscle contraction that is evoked by a suprathreshold stimulation of the muscle or the nerve input.

neurons innervate the same NMJ (muscle fibres could typically receive more than ten different inputs)43, and this polyinnervated stage coincides with the arrival of the first PSCs. Regardless of their number, PSCs are in contact with all terminals and show no apparent prefer-ence22,43. In this situation, PSCs may randomly eliminate branches from different terminals, as suggested by Smith et al.22. Gradually, the number of competing inputs would decrease and the segregation of territories would increase, such that each competing terminal would occupy a pre-cise and delineated territory. When only a few competing terminals remain at an NMJ, PSCs — which can decode the relative synaptic efficacy of different inputs — could trigger the elimination of the weaker inputs. In doing so, PSCs may ensure that the most efficient and suitable nerve terminal is maintained.

Once mono-innervation is achieved, presynaptic and postsynaptic elements continue to rely on PSCs for proper functioning. Indeed, PSCs modulate the fine-tuning of

neurotransmission at mature frog and mouse NMJs153,155 in a Ca2+-dependent manner153,155, whereby the outcome of synaptic regulation is governed by the amplitude and kinetics of intra-PSC Ca2+ signalling. The ablation of PSCs at the mature amphibian NMJ results in structural and functional deficiencies10, including an increase in nerve-terminal retractions, and decreases in both nerve-evoked muscle-twitch tension and the presynaptic release of neuro-transmitter10. Thus, PSCs are important for synapse main-tenance and refinement, as well as for synapse maturation and activity-dependent changes.

Conclusions and perspectivesOwing to its simplicity and accessibility, the NMJ has contributed substantially to our knowledge of synapse formation and maturation. These processes rely on complex molecular interactions that involve synapto-genic factors from muscle fibres, nerve terminals and PSCs. These factors ensure a coordinated maturation

Figure 4 | PSCs can differentiate between competing synaptic terminals. a | A single perisynaptic Schwann cell (PSC; translucent grey) in close association with two terminals (green and blue) during synaptic competition is shown. b,c | PSCs can distinguish between strong (blue) and weak (green) inputs. Neurotransmitter release evoked by selective stimulation of the motor neuron associated with the stronger input induces intra-PSC Ca2+ responses of a greater amplitude (as indicated by the larger percentage change in the calcium-indicator fluorescence over basal fluorescence; %ΔF/F0) than those induced by neurotransmitter release from the weaker input. Insets in part b are false-colour confocal images of changes in intra-PSC Ca2+ level before stimulation (inset 1), at the peak of the response (inset 2) and after stimulation (inset 3). Part c shows intra-PSC Ca2+ responses (mean ± SEM) that were induced by strong and weak competing terminals. Each point represents a single experiment, and the black bars join the intra-PSC Ca2+ responses from the weak and strong nerve terminals that are competing at the same neuromuscular junction (NMJ). | The amplitudes of intra-PSC Ca2+ responses (the PSC-activation index, which is defined as the amplitude of the intra-PSC Ca2+ response induced by the weak input divided by the amplitude of the intra-PSC Ca2+ response induced by the strong input) are tightly correlated with the synaptic strength (the competition index, which is defined as the quantal content of the weak input divided by the quantal content of the strong input and is a major determinant of the outcome of synaptic competition) of competing terminals. Each point is a single experiment, and the black line represents the linear regression relationship (r) between the PSC-activation and synaptic competition indices. Terminals with higher competition and PSC-activation indices are less likely to be eliminated. Parts b–d have been modified and republished with permission of the Society for Neuroscience, from: Glial cells decipher synaptic competition at the mammalian neuromuscular junction. Darabid, H., Arbour, D. & Robitaille, R. J. Neurosci. 33, 1297–1313 (2014); permission conveyed through Copyright Clearance Center, Inc.18.

a

50 Hz, 30 s

1

2

3

1

2

3

b PSC Ca2+ response

1

2

3

50 Hz, 30 s

1

2

3

0

10

20

30

40

50

100

150

Weak input Strong input

PSC

Ca2+

resp

onse

(F/

F0)

0.0

0.5

1.0

0.0 0.5 1.0

P < 0.02r = 0.732

Competition index

PSC

-act

ivat

ion

inde

x

c d

Weak input Strong input

F/F0

F/F0

Muscle bre

Terminal 1 Terminal 2

PSC

Endplate

REVIEWS



of the NMJ such that presynaptic, postsynaptic and glial partners are morphologically matched and func-tionally tuned. This coordination is essential to accu-rately position the hundreds of presynaptic active zones to face the complex postsynaptic organization at the motor endplate. This tight presynaptic–postsynaptic pairing is further compounded by PSC processes, which express receptors that allow them to regulate NMJ efficacy153,155.

The maturation of synapses at the NMJ is influ-enced by changes in synaptic activity and can therefore be tuned as a function of the environment. Moreover, such activity shapes NMJ connectivity by promoting the maintenance of certain inputs and the elimination of redundant connections after competition.

Recent studies have revealed that PSCs regulate many aspects of NMJ formation — from receptor clustering and presynaptic differentiation to activity-dependent synapse pruning and the modulation of synaptic activity. Interestingly, although synapse formation is best studied at the NMJ, the contribution of PSCs to synaptogenesis is less well understood than that of their CNS counter-parts156,157. Indeed, many CNS glial-dependent pathways have been described, including those involving glypicans, thrombospondins, hevin and SPARC (secreted protein acidic and rich in cysteine)158–161. Similar mechanisms are probably involved at the NMJ; however, an assessment of these mechanisms is complicated by the lack of genetic and molecular tools for specifically targeting PSCs.

Therefore, it will be important to develop such tools to unveil the contributions of PSCs and to better understand NMJ formation and maturation.

Furthermore, a functional link between the activity-dependent processes discussed here and the molecular pathways that drive synapse formation and maturation is yet to be elucidated. Considering the high number of intricate interactions among the three elements, it is pos-sible that the molecular mechanisms involved are under some form of activity-dependent regulation. Furthermore, if this is the case, then the activity-dependent regulation of the molecular pathways of each compartment must be accurately controlled to shape a tightly organized com-munication unit. In summary, determining how the dif-ferent molecular and activity-dependent mechanisms influence one another is an important issue that needs to be addressed.

Understanding NMJ formation and the contribution of glia to this process would have a marked impact on our knowledge of the changes that occur at the NMJ with ageing, following nerve injury or in the contexts of vari-ous motor neuron diseases. In particular, understanding the mechanisms that regulate the sprouting and extension of PSC processes would prove to be invaluable, as these are some of the most important events that regulate NMJ reformation following injury. Studying NMJ formation as a tripartite, integrated ensemble process would therefore give us a better understanding of numerous diseases, thus potentially expanding the spectrum of therapeutic targets.

1. Bonanomi, D. & Pfaff, S. L. Motor axon pathfinding. Cold Spring Harb. Perspect. Biol.2, a001735 (2010).

2. O’Donnell, M., Chance, R. K. & Bashaw, G. J. Axon growth and guidance: receptor regulation and signal transduction. Annu. Rev. Neurosci. 32, 383–412 (2009).

3. Desaki, J. & Uehara, Y. The overall morphology of neuromuscular junctions as revealed by scanning electron microscopy. J. Neurocytol. 10, 101–110 (1981).

4. Redfern, P. A. Neuromuscular transmission in new-born rats. J. Physiol. 209, 701–709 (1970).

5. Sanes, J. R. & Lichtman, J. W. Development of the vertebrate neuromuscular junction. Annu. Rev. Neurosci. 22, 389–442 (1999).

6. Patton, B. L. et al. Properly formed but improperly localized synaptic specializations in the absence of laminin α4. Nature Neurosci. 4, 597–604 (2001).

7. Wu, H., Xiong, W. C. & Mei, L. To build a synapse: signaling pathways in neuromuscular junction assembly. Development 137, 1017–1033 (2010).

8. Sanes, J. R. & Lichtman, J. W. Induction, assembly, maturation and maintenance of a postsynaptic apparatus. Nature Rev. Neurosci. 2, 791–805 (2001).

9. Koirala, S., Reddy, L. V. & Ko, C. P. Roles of glial cells in the formation, function, and maintenance of the neuromuscular junction. J. Neurocytol. 32, 987–1002 (2003).

10. Reddy, L. V., Koirala, S., Sugiura, Y., Herrera, A. A. & Ko, C. P. Glial cells maintain synaptic structure and function and promote development of the neuromuscular junction in vivo. Neuron 40, 563–580 (2003).This study demonstrates that selective ablation of PSCs at the developing and mature frog NMJ results in major structural, developmental and functional defects of the synapse.

11. Yang, J.-f. et al. Schwann cells express active agrin and enhance aggregation of acetylcholine receptors on muscle fibers. J. Neurosci. 21, 9572–9584 (2001).

12. Feng, Z. & Ko, C. P. Schwann cells promote synaptogenesis at the neuromuscular junction via

transforming growth factor-β1. J. Neurosci. 28, 9599–9609 (2008).

13. Fuentes-Medel, Y. et al. Integration of a retrograde signal during synapse formation by glia-secreted TGF-β ligand. Curr. Biol. 22, 1831–1838 (2012).

14. Kerr, K. S. et al. Glial Wingless/Wnt regulates glutamate receptor clustering and synaptic physiology at the Drosophila neuromuscular junction. J. Neurosci. 34, 2910–2920 (2014).

15. Escher, P. et al. Synapses form in skeletal muscles lacking neuregulin receptors. Science 308, 1920–1923 (2005).

16. Lin, W. et al. Aberrant development of motor axons and neuromuscular synapses in erbB2-deficient mice. Proc. Natl Acad. Sci. USA 97, 1299–1304 (2000).

17. Wolpowitz, D. et al. Cysteine-rich domain isoforms of the neuregulin-1 gene are required for maintenance of peripheral synapses. Neuron 25, 79–91 (2000).

18. Darabid, H., Arbour, D. & Robitaille, R. Glial cells decipher synaptic competition at the mammalian neuromuscular junction. J. Neurosci. 33, 1297–1313 (2013).By combining confocal calcium-imaging techniques and electrophysiological recordings at dually innervated mouse NMJs, this study shows that PSCs can differentially decode synaptic activity from competing nerve terminals at developing NMJs.

19. Patton, B. L., Chiu, A. Y. & Sanes, J. R. Synaptic laminin prevents glial entry into the synaptic cleft. Nature 393, 698–701 (1998).

20. Yang, D. et al. Coordinate control of axon defasciculation and myelination by laminin-2 and -8. J. Cell Biol. 168, 655–666 (2005).

21. Bishop, D. L., Misgeld, T., Walsh, M. K., Gan, W. B. & Lichtman, J. W. Axon branch removal at developing synapses by axosome shedding. Neuron 44, 651–661 (2004).Using time-lapse imaging and serial electron microscopy of mouse NMJs, this paper describes the changes in the PSC coverage of nerve terminals, providing new insights into the mechanisms of synapse elimination and the shedding of axon terminals by SCs.

22. Smith, I. W., Mikesh, M., Lee, Y. & Thompson, W. J. Terminal Schwann cells participate in the competition underlying neuromuscular synapse elimination. J. Neurosci. 33, 17724–17736 (2013).This research shows that PSCs actively participate in axon removal during synaptic competition in the developing NMJ.

23. Lin, W. et al. Distinct roles of nerve and muscle in postsynaptic differentiation of the neuromuscular synapse. Nature 410, 1057–1064 (2001).

24. Sheard, P. W. & Duxson, M. J. The transient existence of ‘en passant’ nerve terminals in normal embryonic rat skeletal muscle. Brain Res. Dev. Brain Res. 98, 259–264 (1997).

25. Landmesser, L. The development of motor projection patterns in the chick hind limb. J. Physiol. 284, 391–414 (1978).

26. Landmesser, L. The distribution of motoneurones supplying chick hind limb muscles. J. Physiol. 284, 371–389 (1978).

27. Landmesser, L. T. The acquisition of motoneuron subtype identity and motor circuit formation. Int. J. Dev. Neurosci. 19, 175–182 (2001).

28. Jessell, T. M. Neuronal specification in the spinal cord: inductive signals and transcriptional codes. Nature Rev. Genet. 1, 20–29 (2000).

29. Lance-Jones, C. Motoneuron axon guidance: development of specific projections to two muscles in the embryonic chick limb. Brain Behav. Evol. 31, 209–217 (1988).

30. Tripodi, M. & Arber, S. Regulation of motor circuit assembly by spatial and temporal mechanisms. Curr. Opin. Neurobiol. 22, 615–623 (2012).

31. Phelan, K. A. & Hollyday, M. Axon guidance in muscleless chick wings: the role of muscle cells in motoneuronal pathway selection and muscle nerve formation. J. Neurosci. 10, 2699–2716 (1990).

32. Tessier-Lavigne, M. & Goodman, C. S. The molecular biology of axon guidance. Science 274, 1123–1133 (1996).

33. Hanson, M. G. & Landmesser, L. T. Normal patterns of spontaneous activity are required for correct motor axon guidance and the expression of specific guidance molecules. Neuron 43, 687–701 (2004).

REVIEWS



34. Yang, X. et al. Patterning of muscle acetylcholine receptor gene expression in the absence of motor innervation. Neuron 30, 399–410 (2001).

35. Yang, X., Li, W., Prescott, E. D., Burden, S. J. & Wang, J. C. DNA topoisomerase IIβ and neural development. Science 287, 131–134 (2000).

36. Yampolsky, P. et al. Time lapse in vivo visualization of developmental stabilization of synaptic receptors at neuromuscular junctions. J. Biol. Chem. 285, 34589–34596 (2010).

37. Flanagan-Steet, H., Fox, M. A., Meyer, D. & Sanes, J. R. Neuromuscular synapses can form in vivo by incorporation of initially aneural postsynaptic specializations. Development 132, 4471–4481 (2005).

38. Panzer, J. A., Song, Y. & Balice-Gordon, R. J. In vivo imaging of preferential motor axon outgrowth to and synaptogenesis at prepatterned acetylcholine receptor clusters in embryonic zebrafish skeletal muscle. J. Neurosci. 26, 934–947 (2006).

39. Anderson, M. J. & Cohen, M. W. Nerve-induced and spontaneous redistribution of acetylcholine receptors on cultured muscle cells. J. Physiol. 268, 757–773 (1977).

40. Frank, E. & Fischbach, G. D. Early events in neuromuscular junction formation in vitro: induction of acetylcholine receptor clusters in the postsynaptic membrane and morphology of newly formed synapses. J. Cell Biol. 83, 143–158 (1979).

41. Kim, N. & Burden, S. J. MuSK controls where motor axons grow and form synapses. Nature Neurosci. 11, 19–27 (2008).

42. Vock, V. M., Ponomareva, O. N. & Rimer, M. Evidence for muscle-dependent neuromuscular synaptic site determination in mammals. J. Neurosci. 28, 3123–3130 (2008).

43. Tapia, J. C. et al. Pervasive synaptic branch removal in the mammalian neuromuscular system at birth. Neuron 74, 816–829 (2012).

44. Wyatt, R. M. & Balice-Gordon, R. J. Activity-dependent elimination of neuromuscular synapses. J. Neurocytol. 32, 777–794 (2003).

45. Chung, W. S. & Barres, B. A. Selective remodeling: refining neural connectivity at the neuromuscular junction. PLoS Biol. 7, e1000185 (2009).

46. Gautam, M. et al. Defective neuromuscular synaptogenesis in agrin-deficient mutant mice. Cell 85, 525–535 (1996).

47. Bezakova, G., Helm, J. P., Francolini, M. & Lomo, T. Effects of purified recombinant neural and muscle agrin on skeletal muscle fibers in vivo. J. Cell Biol. 153, 1441–1452 (2001).

48. Bezakova, G. & Lomo, T. Muscle activity and muscle agrin regulate the organization of cytoskeletal proteins and attached acetylcholine receptor (AchR) aggregates in skeletal muscle fibers. J. Cell Biol. 153, 1453–1463 (2001).

49. Jones, G. et al. Induction by agrin of ectopic and functional postsynaptic-like membrane in innervated muscle. Proc. Natl Acad. Sci. USA 94, 2654–2659 (1997).

50. Gesemann, M., Denzer, A. J. & Ruegg, M. A. Acetylcholine receptor-aggregating activity of agrin isoforms and mapping of the active site. J. Cell Biol. 128, 625–636 (1995).

51. Burgess, R. W., Nguyen, Q. T., Son, Y. J., Lichtman, J. W. & Sanes, J. R. Alternatively spliced isoforms of nerve- and muscle-derived agrin: their roles at the neuromuscular junction. Neuron 23, 33–44 (1999).

52. Bowen, D. C., Sugiyama, J., Ferns, M. & Hall, Z. W. Neural agrin activates a high-affinity receptor in C2 muscle cells that is unresponsive to muscle agrin. J. Neurosci. 16, 3791–3797 (1996).

53. Ferns, M. J., Campanelli, J. T., Hoch, W., Scheller, R. H. & Hall, Z. The ability of agrin to cluster AChRs depends on alternative splicing and on cell surface proteoglycans. Neuron 11, 491–502 (1993).

54. DeChiara, T. M. et al. The receptor tyrosine kinase MuSK is required for neuromuscular junction formation in vivo. Cell 85, 501–512 (1996).

55. Valenzuela, D. M. et al. Receptor tyrosine kinase specific for the skeletal muscle lineage: expression in embryonic muscle, at the neuromuscular junction, and after injury. Neuron 15, 573–584 (1995).

56. Weatherbee, S. D., Anderson, K. V. & Niswander, L. A. LDL-receptor-related protein 4 is crucial for formation of the neuromuscular junction. Development 133, 4993–5000 (2006).

57. Zhang, B. et al. LRP4 serves as a coreceptor of agrin. Neuron 60, 285–297 (2008).

58. Kim, N. et al. Lrp4 is a receptor for Agrin and forms a complex with MuSK. Cell 135, 334–342 (2008).This is the first study to demonstrate that LRP4 is the agrin co-receptor that is required for MUSK activation.

59. Burden, S. J., Yumoto, N. & Zhang, W. The role of MuSK in synapse formation and neuromuscular disease. Cold Spring Harb. Perspect. Biol. 5, a009167 (2013).

60. Gautam, M. et al. Failure of postsynaptic specialization to develop at neuromuscular junctions of rapsyn-deficient mice. Nature 377, 232–236 (1995).

61. Zhang, B. et al. Wnt proteins regulate acetylcholine receptor clustering in muscle cells. Mol. Brain 5, 7 (2012).

62. Korkut, C. & Budnik, V. WNTs tune up the neuromuscular junction. Nature Rev. Neurosci. 10, 627–634 (2009).

63. Budnik, V. & Salinas, P. C. Wnt signaling during synaptic development and plasticity. Curr. Opin. Neurobiol. 21, 151–159 (2011).

64. Jing, L., Lefebvre, J. L., Gordon, L. R. & Granato, M. Wnt signals organize synaptic prepattern and axon guidance through the zebrafish unplugged/MuSK receptor. Neuron 61, 721–733 (2009).

65. Henriquez, J. P. et al. Wnt signaling promotes AChR aggregation at the neuromuscular synapse in collaboration with agrin. Proc. Natl Acad. Sci. USA 105, 18812–18817 (2008).

66. Wang, J. et al. Wnt/β-catenin signaling suppresses rapsyn expression and inhibits acetylcholine receptor clustering at the neuromuscular junction. J. Biol. Chem. 283, 21668–21675 (2008).

67. Castelo-Branco, G. et al. Ventral midbrain glia express region-specific transcription factors and regulate dopaminergic neurogenesis through Wnt-5a secretion. Mol. Cell Neurosci. 31, 251–262 (2006).

68. Rimer, M., Cohen, I., Lomo, T., Burden, S. J. & McMahan, U. J. Neuregulins and erbB receptors at neuromuscular junctions and at agrin-induced postsynaptic-like apparatus in skeletal muscle. Mol. Cell Neurosci. 12, 1–15 (1998).

69. Lemke, G. Glial control of neuronal development. Annu. Rev. Neurosci. 24, 87–105 (2001).

70. Jo, S. A., Zhu, X., Marchionni, M. A. & Burden, S. J. Neuregulins are concentrated at nerve–muscle synapses and activate ACh-receptor gene expression. Nature 373, 158–161 (1995).

71. Sandrock, A. W. Jr et al. Maintenance of acetylcholine receptor number by neuregulins at the neuromuscular junction in vivo. Science 276, 599–603 (1997).

72. Riethmacher, D. et al. Severe neuropathies in mice with targeted mutations in the ErbB3 receptor. Nature 389, 725–730 (1997).

73. Ngo, S. T., Cole, R. N., Sunn, N., Phillips, W. D. & Noakes, P. G. Neuregulin-1 potentiates agrin-induced acetylcholine receptor clustering through muscle-specific kinase phosphorylation. J. Cell Sci. 125, 1531–1543 (2012).

74. Jaworski, A. & Burden, S. J. Neuromuscular synapse formation in mice lacking motor neuron- and skeletal muscle-derived neuregulin-1. J. Neurosci. 26, 655–661 (2006).

75. Ponomareva, O. N. et al. Defective neuromuscular synaptogenesis in mice expressing constitutively active ErbB2 in skeletal muscle fibers. Mol. Cell Neurosci. 31, 334–345 (2006).

76. Garratt, A. N., Britsch, S. & Birchmeier, C. Neuregulin, a factor with many functions in the life of a schwann cell. Bioessays 22, 987–996 (2000).

77. St John, P. A. & Gordon, H. Agonists cause endocytosis of nicotinic acetylcholine receptors on cultured myotubes. J. Neurobiol. 49, 212–223 (2001).

78. Misgeld, T., Kummer, T. T., Lichtman, J. W. & Sanes, J. R. Agrin promotes synaptic differentiation by counteracting an inhibitory effect of neurotransmitter. Proc. Natl Acad. Sci. USA 102, 11088–11093 (2005).This elegant study uses double-mutant mice (lacking both ChAT and agrin) to show that synaptic transmission regulates NMJ maturation and counters the effects of agrin.

79. Misgeld, T. et al. Roles of neurotransmitter in synapse formation: development of neuromuscular junctions lacking choline acetyltransferase. Neuron 36, 635–648 (2002).

80. Brandon, E. P. et al. Aberrant patterning of neuromuscular synapses in choline acetyltransferase-deficient mice. J. Neurosci. 23, 539–549 (2003).

81. Singhal, N. & Martin, P. T. Role of extracellular matrix proteins and their receptors in the development of the vertebrate neuromuscular junction. Dev. Neurobiol. 71, 982–1005 (2011).

82. VanSaun, M., Herrera, A. A. & Werle, M. J. Structural alterations at the neuromuscular junctions of matrix metalloproteinase 3 null mutant mice. J. Neurocytol. 32, 1129–1142 (2003).

83. VanSaun, M. & Werle, M. J. Matrix metalloproteinase-3 removes agrin from synaptic basal lamina. J. Neurobiol. 44, 369 (2000).

84. Werle, M. J. Cell-to-cell signaling at the neuromuscular junction: the dynamic role of the extracellular matrix. Ann. NY Acad. Sci. 1132, 13–18 (2008).

85. Werle, M. J. & VanSaun, M. Activity dependent removal of agrin from synaptic basal lamina by matrix metalloproteinase 3. J. Neurocytol. 32, 905–913 (2003).

86. Nagase, H. & Woessner, J. F. Jr. Matrix metalloproteinases. J. Biol. Chem. 274, 21491–21494 (1999).

87. Knight, D., Tolley, L. K., Kim, D. K., Lavidis, N. A. & Noakes, P. G. Functional analysis of neurotransmission at β2-laminin deficient terminals. J. Physiol. 546, 789–800 (2003).

88. Noakes, P. G., Gautam, M., Mudd, J., Sanes, J. R. & Merlie, J. P. Aberrant differentiation of neuromuscular junctions in mice lacking s-laminin/laminin β 2. Nature 374, 258–262 (1995).

89. Nishimune, H., Sanes, J. R. & Carlson, S. S. A synaptic laminin–calcium channel interaction organizes active zones in motor nerve terminals. Nature 432, 580–587 (2004).

90. Urbano, F. J., Rosato-Siri, M. D. & Uchitel, O. D. Calcium channels involved in neurotransmitter release at adult, neonatal and P/Q-type deficient neuromuscular junctions (review). Mol. Membr. Biol. 19, 293–300 (2002).

91. Chen, J., Billings, S. E. & Nishimune, H. Calcium channels link the muscle-derived synapse organizer laminin β2 to Bassoon and CAST/Erc2 to organize presynaptic active zones. J. Neurosci. 31, 512–525 (2011).

92. Umemori, H., Linhoff, M. W., Ornitz, D. M. & Sanes, J. R. FGF22 and its close relatives are presynaptic organizing molecules in the mammalian brain. Cell 118, 257–270 (2004).

93. Fox, M. A. et al. Distinct target-derived signals organize formation, maturation, and maintenance of motor nerve terminals. Cell 129, 179–193 (2007).This elegant study uncovers the role of FGFs, laminin β2 and type IV collagen α chains using specific in vivo genetic manipulations to describe their sequential role in presynaptic maturation.

94. Zhang, X. et al. Receptor specificity of the fibroblast growth factor family. The complete mammalian FGF family. J. Biol. Chem. 281, 15694–15700 (2006).

95. Miner, J. H. & Sanes, J. R. Collagen IV α 3, α 4, and α 5 chains in rodent basal laminae: sequence, distribution, association with laminins, and developmental switches. J. Cell Biol. 127, 879–891 (1994).

96. Latvanlehto, A. et al. Muscle-derived collagen XIII regulates maturation of the skeletal neuromuscular junction. J. Neurosci. 30, 12230–12241 (2010).

97. Umemori, H. & Sanes, J. R. Signal regulatory proteins (SIRPS) are secreted presynaptic organizing molecules. J. Biol. Chem. 283, 34053–34061 (2008).

98. Henderson, C. E. et al. GDNF: a potent survival factor for motoneurons present in peripheral nerve and muscle. Science 266, 1062–1064 (1994).

99. Wright, D. E. & Snider, W. D. Focal expression of glial cell line-derived neurotrophic factor in developing mouse limb bud. Cell Tissue Res. 286, 209–217 (1996).

100. Yan, Q., Matheson, C. & Lopez, O. T. In vivo neurotrophic effects of GDNF on neonatal and adult facial motor neurons. Nature 373, 341–344 (1995).

101. Nguyen, Q. T. Hyperinnervation of neuromuscular junctions caused by GDNF overexpression in muscle. Science 279, 1725–1729 (1998).

102. Je, H. S. et al. ProBDNF and mature BDNF as punishment and reward signals for synapse elimination at mouse neuromuscular junctions. J. Neurosci. 33, 9957–9962 (2013).

103. Chao, M. V. & Bothwell, M. Neurotrophins: to cleave or not to cleave. Neuron 33, 9–12 (2002).

104. Je, H. S. et al. Role of pro-brain-derived neurotrophic factor (proBDNF) to mature BDNF conversion in activity-dependent competition at developing neuromuscular synapses. Proc. Natl Acad. Sci. USA 109, 15924–15929 (2012).

105. Nishimune, H. et al. Laminins promote postsynaptic maturation by an autocrine mechanism at the neuromuscular junction. J. Cell Biol. 182, 1201–1215 (2008).

REVIEWS



106. Jacobson, C., Cote, P. D., Rossi, S. G., Rotundo, R. L. & Carbonetto, S. The dystroglycan complex is necessary for stabilization of acetylcholine receptor clusters at neuromuscular junctions and formation of the synaptic basement membrane. J. Cell Biol. 152, 435–450 (2001).

107. Cote, P. D., Moukhles, H., Lindenbaum, M. & Carbonetto, S. Chimaeric mice deficient in dystroglycans develop muscular dystrophy and have disrupted myoneural synapses. Nature Genet. 23, 338–342 (1999).

108. Arikawa-Hirasawa, E., Rossi, S. G., Rotundo, R. L. & Yamada, Y. Absence of acetylcholinesterase at the neuromuscular junctions of perlecan-null mice. Nature Neurosci. 5, 119–123 (2002).

109. Zhang, W., Coldefy, A. S., Hubbard, S. R. & Burden, S. J. Agrin binds to the N-terminal region of Lrp4 protein and stimulates association between Lrp4 and the first immunoglobulin-like domain in muscle-specific kinase (MuSK). J. Biol. Chem. 286, 40624–40630 (2011).

110. Yumoto, N., Kim, N. & Burden, S. J. Lrp4 is a retrograde signal for presynaptic differentiation at neuromuscular synapses. Nature 489, 438–442 (2012).

111. Huijbers, M. G. et al. MuSK IgG4 autoantibodies cause myasthenia gravis by inhibiting binding between MuSK and Lrp4. Proc. Natl Acad. Sci. USA 110, 20783–20788 (2013).

112. Wu, H. et al. Distinct roles of muscle and motoneuron LRP4 in neuromuscular junction formation. Neuron 75, 94–107 (2012).This study describes, using cell-specific mutations, a potential role of motor neuron LRP4, in addition to muscular LRP4, in the maturation of the NMJ.

113. Packard, M. et al. The Drosophila Wnt, wingless, provides an essential signal for pre- and postsynaptic differentiation. Cell 111, 319–330 (2002).

114. Liebl, F. L. et al. Derailed regulates development of the Drosophila neuromuscular junction. Dev. Neurobiol. 68, 152–165 (2008).

115. Wells, D. G., McKechnie, B. A., Kelkar, S. & Fallon, J. R. Neurotrophins regulate agrin-induced postsyn aptic differentiation. Proc. Natl Acad. Sci. USA 96, 1112–1117 (1999).

116. Shelton, D. L. et al. Human TRKs: molecular cloning, tissue distribution, and expression of extracellular domain immunoadhesins. J. Neurosci. 15, 477–491 (1995).

117. Gonzalez, M. et al. Disruption of Trkb-mediated signaling induces disassembly of postsynaptic receptor clusters at neuromuscular junctions. Neuron 24, 567–583 (1999).

118. Yang, L. X. & Nelson, P. G. Glia cell line-derived neurotrophic factor regulates the distribution of acetylcholine receptors in mouse primary skeletal muscle cells. Neuroscience 128, 497–509 (2004).

119. Todd, K. J., Auld, D. S. & Robitaille, R. Neurotrophins modulate neuron-glia interactions at a vertebrate synapse. Eur. J. Neurosci. 25, 1287–1296 (2007).

120. Mirsky, R. et al. Schwann cells as regulators of nerve development. J. Physiol. Paris 96, 17–24 (2002).

121. Davies, A. M. Neuronal survival: early dependence on Schwann cells. Curr. Biol. 8, R15–R18 (1998).

122. Halstead, S. K. et al. Anti-disialosyl antibodies mediate selective neuronal or Schwann cell injury at mouse neuromuscular junctions. Glia 52, 177–189 (2005).

123. Halstead, S. K. et al. Anti-disialoside antibodies kill perisynaptic Schwann cells and damage motor nerve terminals via membrane attack complex in a murine model of neuropathy. Brain 127, 2109–2123 (2004).

124. Balic e-Gordon, R. J., Chua, C. K., Nelson, C. C. & Lichtman, J. W. Gradual loss of synaptic cartels precedes axon withdrawal at developing neuromuscular junctions. Neuron 11, 801–815 (1993).

125. Walsh, M. K. & Lichtman, J. W. In vivo time-lapse imaging of synaptic takeover associated with naturally occurring synapse elimination. Neuron 37, 67–73 (2003).This is a detailed analysis using confocal and in vivo imaging that describes the morphological changes that take place in the NMJ during synaptic competition and elimination in development.

126. Cai, D., Cohen, K. B., Luo, T., Lichtman, J. W. & Sanes, J. R. Improved tools for the Brainbow toolbox. Nature Methods 10, 540–547 (2013).

127. Lu, J., Tapia, J. C., White, O. L. & Lichtman, J. W. The interscutularis muscle connectome. PLoS Biol. 7, e32 (2009).

128. Gan, W. B. & Lichtman, J. W. Synaptic segregation at the developing neuromuscular junction. Science 282, 1508–1511 (1998).

129. Turney, S. G. & Lichtman, J. W. Reversing the outcome of synapse elimination at developing neuromuscular junctions in vivo: evidence for synaptic competition and its mechanism. PLoS Biol. 10, e1001352 (2012).Using specific ablation of individual nerve terminals during NMJ formation, this study shows that active nerve terminals can innervate vacated territories, underlining the notion that competition at developing NMJs is an active process.

130. Colman, H., Nabekura, J. & Lichtman, J. W. Alterations in synaptic strength preceding axon withdrawal. Science 275, 356–361 (1997).

131. Buffelli, M. et al. Genetic evidence that relative synaptic efficacy biases the outcome of synaptic competition. Nature 424, 430–434 (2003).This study uses genetic tools to show that inputs that release more neurotransmitter are favoured during synapse elimination at the developing NMJ.

132. Kopp, D. M., Perkel, D. J. & Balice-Gordon, R. J. Disparity in neurotransmitter release probability among competing inputs during neuromuscular synapse elimination. J. Neurosci. 20, 8771–8779 (2000).

133. Personius, K. E. & Balice-Gordon, R. J. Book review: activity-dependent synaptic plasticity: insights from neuromuscular junctions. Neuroscientist 8, 414–422 (2002).

134. Callaway, E. M., Soha, J. M. & Van Essen, D. C. Competition favouring inactive over active motor neurons during synapse elimination. Nature 328, 422–426 (1987).

135. Costanzo, E. M., Barry, J. A. & Ribchester, R. R. Co-regulation of synaptic efficacy at stable polyneuronally innervated neuromuscular junctions in reinnervated rat muscle. J. Physiol. 521 (Pt. 2), 365–374 (1999).

136. Costanzo, E. M., Barry, J. A. & Ribchester, R. R. Competition at silent synapses in reinnervated skeletal muscle. Nature Neurosci. 3, 694–700 (2000).

137. Favero, M., Busetto, G. & Cangiano, A. Spike timing plays a key role in synapse elimination at the neuromuscular junction. Proc. Natl Acad. Sci. USA 109, E1667–E1675 (2012).

138. Balice-Gordon, R. J. & Lichtman, J. W. Long-term synapse loss induced by focal blockade of postsynaptic receptors. Nature 372, 519–524 (1994).

139. Personius, K. E. & Balice-Gordon, R. J. Loss of correlated motor neuron activity during synaptic competition at developing neuromuscular synapses. Neuron 31, 395–408 (2001).This study uses in vivo electrophysiology techniques to study the patterns of changes in the activity of motor neurons during synaptic competition and elimination.

140. Ribchester, R. R. & Taxt, T. Motor unit size and synaptic competition in rat lumbrical muscles reinnervated by active and inactive motor axons. J. Physiol. 344, 89–111 (1983).

141. Ridge, R. M. & Betz, W. J. The effect of selective, chronic stimulation on motor unit size in developing rat muscle. J. Neurosci. 4, 2614–2620 (1984).

142. Buffelli, M., Busetto, G., Cangiano, L. & Cangiano, A. Perinatal switch from synchronous to asynchronous activity of motoneurons: link with synapse elimination. Proc. Natl Acad. Sci. USA 99, 13200–13205 (2002).

143. Marques, M. J., Conchello, J. A. & Lichtman, J. W. From plaque to pretzel: fold formation and acetylcholine receptor loss at the developing neuromuscular junction. J. Neurosci. 20, 3663–3675 (2000).

144. Neukomm, L. J. & Freeman, M. R. Diverse cellular and molecular modes of axon degeneration. Trends Cell Biol. 24, 1–9 (2014).

145. Pfrieger, F. W. Roles of glial cells in synapse development. Cell. Mol. Life Sci. 66, 2037–2047 (2009).

146. Song, J. W. et al. Lysosomal activity associated with developmental axon pruning. J. Neurosci. 28, 8993–9001 (2008).

147. Corty, M. M. & Freeman, M. R. Cell biology in neuroscience: architects in neural circuit design: glia control neuron numbers and connectivity. J. Cell Biol. 203, 395–405 (2013).

148. Fuentes-Medel, Y. et al. Glia and muscle sculpt neuromuscular arbors by engulfing destabilized synaptic boutons and shed presynaptic debris. PLoS Biol. 7, e1000184 (2009).

149. Brill, M. S., Lichtman, J. W., Thompson, W., Zuo, Y. & Misgeld, T. Spatial constraints dictate glial territories at murine neuromuscular junctions. J. Cell Biol. 195, 293–305 (2011).This paper shows that following the ablation of individual SCs, changes in PSC-territory dynamics at developing and mature NMJs are under spatial constraints.

150. Hirata, K., Zhou, C., Nakamura, K. & Kawabuchi, M. Postnatal development of Schwann cells at neuromuscular junctions, with special reference to synapse elimination. J. Neurocytol. 26, 799–809 (1997).

151. Love, F. M. & Thompson, W. J. Schwann cells proliferate at rat neuromuscular junctions during development and regeneration. J. Neurosci. 18, 9376–9385 (1998).

152. Jahromi, B. S., Robitaille, R. & Charlton, M. P. Transmitter release increases intracellular calcium in perisynaptic Schwann cells in situ. Neuron 8, 1069–1077 (1992).

153. Robitaille, R. Modulation of synaptic efficacy and synaptic depression by glial cells at the frog neuromuscular junction. Neuron 21, 847–855 (1998).

154. Rochon, D., Rousse, I. & Robitaille, R. Synapse–glia interactions at the mammalian neuromuscular junction. J. Neurosci. 21, 3819–3829 (2001).

155. Todd, K. J., Darabid, H. & Robitaille, R. Perisynaptic glia discriminate patterns of motor nerve activity and influence plasticity at the neuromuscular junction. J. Neurosci. 30, 11870–11882 (2010).This study combines electrophysiological recordings, PSC Ca2+-imaging and alteration of the activity of specific PSCs using photo-activatable molecules to show that PSCs regulate synaptic plasticity by discerning patterns of motor nerve activity.

156. Clarke, L. E. & Barres, B. A. Emerging roles of astrocytes in neural circuit development. Nature Rev. Neurosci. 14, 311–321 (2013).

157. Eroglu, C. & Barres, B. A. Regulation of synaptic connectivity by glia. Nature 468, 223–231 (2010).

158. Allen, N. J. et al. Astrocyte glypicans 4 and 6 promote formation of excitatory synapses via GluA1 AMPA receptors. Nature 486, 410–414 (2012).

159. Christopherson, K. S. et al. Thrombospondins are astrocyte-secreted proteins that promote CNS synaptogenesis. Cell 120, 421–433 (2005).

160. Kucukdereli, H. et al. Control of excitatory CNS synaptogenesis by astrocyte-secreted proteins Hevin and SPARC. Proc. Natl Acad. Sci. USA 108, E440–449 (2011).

161. Murai, K. K. & Pasquale, E. B. Eph receptors and ephrins in neuron-astrocyte communication at synapses. Glia 59, 1567–1578 (2011).

162. Jessen, K. R. & Mirsky, R. The origin and development of glial cells in peripheral nerves. Nature Rev. Neurosci. 6, 671–682 (2005).

163. Georgiou, J., Robitaille, R. & Charlton, M. P. Muscarinic control of cytoskeleton in perisynaptic glia. J. Neurosci. 19, 3836–3846 (1999).

164. Fawcett, J. W. & Keynes, R. J. Peripheral nerve regeneration. Annu. Rev. Neurosci. 13, 43–60 (1990).

165. Kang, H., Tian, L. & Thompson, W. Terminal Schwann cells guide the reinnervation of muscle after nerve injury. J. Neurocytol 32, 975–985 (2003).

166. Miledi, R. & Slater, C. R. Electrophysiology and electron-microscopy of rat neuromuscular junctions after nerve degeneration. Proc. R. Soc. Lond. B Biol. Sci. 169, 289–306 (1968).

167. Miledi, R. & Slater, C. R. On the degeneration of rat neuromuscular junctions after nerve section. J. Physiol. 207, 507–528 (1970).

168. Reynolds, M. L. & Woolf, C. J. Terminal Schwann cells elaborate extensive processes following denervation of the motor endplate. J. Neurocytol. 21, 50–66 (1992).

169. O’Malley, J. P., Waran, M. T. & Balice-Gordon, R. J. In vivo observations of terminal Schwann cells at normal, denervated, and reinnervated mouse neuromuscular junctions. J. Neurobiol. 38, 270–286 (1999).

170. Gantus, M. A., Nasciutti, L. E., Cruz, C. M., Persechini, P. M. & Martinez, A. M. Modulation of extracellular matrix components by metalloproteinases and their tissue inhibitors during degeneration and regeneration of rat sural nerve. Brain Res. 1122, 36–46 (2006).

171. Rich, M. M. & Lichtman, J. W. In vivo visualization of pre- and postsynaptic changes during synapse elimination in reinnervated mouse muscle. J. Neurosci. 9, 1781–1805 (1989).

REVIEWS



172. Kang, H. & Lichtman, J. W. Motor axon regeneration and muscle reinnervation in young adult and aged animals. J. Neurosci. 33, 19480–19491 (2013).

173. Magill, C. K. et al. Reinnervation of the tibialis anterior following sciatic nerve crush injury: a confocal microscopic study in transgenic mice. Exp. Neurol. 207, 64–74 (2007).

174. Laskowski, M. B., Colman, H., Nelson, C. & Lichtman, J. W. Synaptic competition during the reformation of a neuromuscular map. J. Neurosci. 18, 7328–7335 (1998).

175. Lindstrom, J. M., Seybold, M. E., Lennon, V. A., Whittingham, S. & Duane, D. D. Antibody to acetylcholine receptor in myasthenia gravis. Prevalence, clinical correlates, and diagnostic value. Neurology 26, 1054–1059 (1976).

176. Hoch, W. et al. Auto-antibodies to the receptor tyrosine kinase MuSK in patients with myasthenia gravis without acetylcholine receptor antibodies. Nature Med. 7, 365–368 (2001).

177. Plomp, J. J., Huijbers, M. G., van der Maarel, S. M. & Verschuuren, J. J. Pathogenic IgG4 subclass autoantibodies in MuSK myasthenia gravis. Ann. NY Acad. Sci. 1275, 114–122 (2012).

178. Mori, S. et al. Antibodies against muscle-specific kinase impair both presynaptic and postsynaptic

functions in a murine model of myasthenia gravis. Am. J. Pathol. 180, 798–810 (2012).

179. Higuchi, O., Hamuro, J., Motomura, M. & Yamanashi, Y. Autoantibodies to low-density lipoprotein receptor-related protein 4 in myasthenia gravis. Ann. Neurol. 69, 418–422 (2011).

180. Shen, C. et al. Antibodies against low-density lipoprotein receptor-related protein 4 induce myasthenia gravis. J. Clin. Invest. 123, 5190–5202 (2013).

181. Huijbers, M. G. et al. Pathogenic immune mechanisms at the neuromuscular synapse: the role of specific antibody-binding epitopes in myasthenia gravis. J. Intern. Med. 275, 12–26 (2014).

182. Maselli, R. A., Arredondo, J., Ferns, M. J. & Wollmann, R. L. Synaptic basal lamina-associated congenital myasthenic syndromes. Ann. NY Acad. Sci. 1275, 36–48 (2012).

183. Musarella, M. et al. Expression of Nav1.6 sodium channels by Schwann cells at neuromuscular junctions: role in the motor endplate disease phenotype. Glia 53, 13–23 (2006).

184. Caillol, G. et al. Motor endplate disease affects neuromuscular junction maturation. Eur. J. Neurosci. 36, 2400–2408 (2012).

185. Lin, W. et al. Neurotransmitter acetylcholine negatively regulates neuromuscular synapse formation

by a Cdk5-dependent mechanism. Neuron 46, 569–579 (2005).

186. Son, Y. J. et al. The synaptic vesicle protein SV2 is complexed with an α 5-containing laminin on the nerve terminal surface. J. Biol. Chem. 275, 451–460 (2000).

187. Xu, P. et al. Nerve injury induces glial cell line-derived neurotrophic factor (GDNF) expression in Schwann cells through purinergic signaling and the PKC–PKD pathway. Glia 61, 1029–1040 (2013).

AcknowledgementsThe authors thank S. Carbonetto and A. Llobet for reading the manuscript and providing useful comments. This work was sup-ported by grants from the Canadian Institutes for Health Research to R.R. (MOP-14137 and MOP-111070), a Leader Opportunity Fund from the Canadian Foundation of Innovation and an infrastructure grant from Fonds Recherche Québec-Santé (FRQ-S) to the GRSNC (Groupe de Recherche sur le Système Nerveux Central). At the time of writing, H.D. held a FRQ-S studentship.

Competing interests statementThe authors declare no competing interests.

REVIEWS



neuromuscular synaptogenesis: coordinating partners...

Documents